Engineering Papers⌕ Search

SEARCH · Engineering Papers

Results for “distributed memory parallel algorithm”

Search indexed NASA NTRS and DOE OSTI research on propulsion, heat transfer, battery materials and energy systems. Follow report and document links to the original sources.

Quote a phrase for an exact phrase match. Source license links do not imply unrestricted reuse.

At least 361 records · Page 20

Mobile and replicated alignment of arrays in data-parallel programs

When a data-parallel language like FORTRAN 90 is compiled for a distributed-memory machine, aggregate data objects (such as arrays) are distributed across the processor memories. The mapping determines the amount of residual communication needed to bring operands of parallel operations into alignment with each other. A common approach is to break the mapping into two stages: first, an alignment that maps all the objects to an abstract template, and then a distribution that maps the template to the processors. We solve two facets of the problem of finding alignments that reduce residual communication: we determine alignments that vary in loops, and objects that should have replicated alignments. We show that loop-dependent mobile alignment is sometimes necessary for optimum performance, and we provide algorithms with which a compiler can determine good mobile alignments for objects within do loops. We also identify situations in which replicated alignment is either required by the program itself (via spread operations) or can be used to improve performance. We propose an algorithm based on network flow that determines which objects to replicate so as to minimize the total amount of broadcast communication in replication. This work on mobile and replicated alignment extends our earlier work on determining static alignment.

Chatterjee, Siddhartha↗

A performance model for the communication in fast multipole methods on high-performance computing platforms

Exascale systems are predicted to have approximately 1 billion cores, assuming gigahertz cores. Limitations on affordable network topologies for distributed memory systems of such massive scale bring new challenges to the currently dominant parallel programing model. Currently, there are many efforts to evaluate the hardware and software bottlenecks of exascale designs. It is therefore of interest to model application performance and to understand what changes need to be made to ensure extrapolated scalability. The fast multipole method (FMM) was originally developed for accelerating N-body problems in astrophysics and molecular dynamics but has recently been extended to a wider range of problems. Its high arithmetic intensity combined with its linear complexity and asynchronous communication patterns make it a promising algorithm for exascale systems. In this paper, we discuss the challenges for FMM on current parallel computers and future exascale architectures, with a focus on internode communication. We focus on the communication part only; the efficiency of the computational kernels are beyond the scope of the present study. We develop a performance model that considers the communication patterns of the FMM and observe a good match between our model and the actual communication time on four high-performance computing (HPC) systems, when latency, bandwidth, network topology, and multicore penalties are all taken into account. To our knowledge, this is the first formal characterization of internode communication in FMM that validates the model against actual measurements of communication time. The ultimate communication model is predictive in an absolute sense; however, on complex systems, this objective is often out of reach or of a difficulty out of proportion to its benefit when there exists a simpler model that is inexpensive and sufficient to guide coding decisions leading to improved scaling. The current model provides such guidance.

Computer Science↗

Productive Programming of Distributed Systems with the SHAD C++ Library

High-performance computing (HPC) is often perceived as a matter of making large-scale systems (e.g., clusters) run as fast as possible, regardless the required programming effort. However, the idea of "bringing HPC to the masses" has recently emerged. Inspired by this vision, we have designed SHAD, the Scalable High-performance Algorithms and Data-structures library. SHAD is open source software, written in C++, for C++ developers. Unlike other HPC libraries for distributed systems, which rely on SPMD models, SHAD adopts a shared-memory programming abstraction, to make C++ programmers feel at home. Underneath, SHAD manages tasking and data-movements, moving the computation where data resides and taking advantage of asynchrony to tolerate network latency. At the bottom of his stack, SHAD can interface with multiple runtime systems: this not only improves developer’s productivity, by hiding the complexity of such software and of the underlying hardware, but also greatly enhance code portability. Thanks to its abstraction layers, SHAD can indeed target different systems, ranging from laptops to HPC clusters, without any need for modifying the user-level code. We have prototyped and open-sourced the implementation of (a subset of) the C++ standard library (STL) targeting multi-node HPC clusters. Our work allows plain STL-based C++ code to scale on HPC systems, with no need for rewriting the code to exploit the complex hardware. SHAD is available under Apache v2 License at https://github.com/pnnl/SHAD. In this paper we overview the design of the SHAD library, depicting its main components: runtime systems abstractions for tasking; parallel and distributed data-structures; STL-compliant interfaces and algorithms.

Castellana, Vito G.↗

Parallel Implicit Algorithms for CFD

The main goal of this project was efficient distributed parallel and workstation cluster implementations of Newton-Krylov-Schwarz (NKS) solvers for implicit Computational Fluid Dynamics (CFD.) "Newton" refers to a quadratically convergent nonlinear iteration using gradient information based on the true residual, "Krylov" to an inner linear iteration that accesses the Jacobian matrix only through highly parallelizable sparse matrix-vector products, and "Schwarz" to a domain decomposition form of preconditioning the inner Krylov iterations with primarily neighbor-only exchange of data between the processors. Prior experience has established that Newton-Krylov methods are competitive solvers in the CFD context and that Krylov-Schwarz methods port well to distributed memory computers. The combination of the techniques into Newton-Krylov-Schwarz was implemented on 2D and 3D unstructured Euler codes on the parallel testbeds that used to be at LaRC and on several other parallel computers operated by other agencies or made available by the vendors. Early implementations were made directly in Massively Parallel Integration (MPI) with parallel solvers we adapted from legacy NASA codes and enhanced for full NKS functionality. Later implementations were made in the framework of the PETSC library from Argonne National Laboratory, which now includes pseudo-transient continuation Newton-Krylov-Schwarz solver capability (as a result of demands we made upon PETSC during our early porting experiences). A secondary project pursued with funding from this contract was parallel implicit solvers in acoustics, specifically in the Helmholtz formulation. A 2D acoustic inverse problem has been solved in parallel within the PETSC framework.

Keyes, David E.↗

A scalable approach to solving dense linear algebra problems on hybrid CPU-GPU systems

Aiming to fully exploit the computing power of all CPUs and all graphics processing units (GPUs) on hybrid CPU-GPU systems to solve dense linear algebra problems, in this paper we design a class of heterogeneous tile algorithms to maximize the degree of parallelism, to minimize the communication volume, and to accommodate the heterogeneity between CPUs and GPUs. The new heterogeneous tile algorithms are executed upon our decentralized dynamic scheduling runtime system, which schedules a task graph dynamically and transfers data between compute nodes automatically. The runtime system uses a new distributed task assignment protocol to solve data dependencies between tasks without any coordination between processing units. By overlapping computation and communication through dynamic scheduling, we are able to attain scalable performance for the double-precision Cholesky factorization and QR factorization. Finally, our approach demonstrates a performance comparable to Intel MKL on shared-memory multicore systems and better performance than both vendor (e.g., Intel MKL) and open source libraries (e.g., StarPU) in the following three environments: heterogeneous clusters with GPUs, conventional clusters without GPUs, and shared-memory systems with multiple GPUs.

97 MATHEMATICS AND COMPUTING↗

On-the-fly Doppler Broadening with Probability Table Interpolation for Unresolved Resonance Region in RMC

With the rapid development of computational power and parallel algorithms, Monte Carlo method has been widely investigated and used in neutron transport simulations due to its advantages in geometry modelling and usage of continuous energy point-wise cross sections. Nowadays the road has been paved to couple Monte Carlo codes with thermal-hydraulics codes to make multi-physics calculations. One of the most important aspects of thermal-hydraulics feedback is the detailed temperature distribution, which demands temperature dependent cross sections in neutron transport simulations. The conventional pre-generated cross section libraries at each temperature point are no longer feasible for detailed temperature distributions because of tremendous memory footprint. Hence, different on-the-fly Doppler broadening methods have been proposed for resolved resonance energy region and thermal energy region, such as the interpolation, regress model fitting, multipole representation (MPR), target motion sampling (TMS) and different integration algorithms based on Doppler broadening equation. Compared to resolved resonance energy region and thermal energy region, the representation of cross sections in unresolved resonance region is slightly different. As a result of the limitation of experiment measurements, the overlapped resonances cannot be distinguished from each other and the statistic average resonance parameters are given in the evaluated nuclear data file. In early applications, infinite-dilute pointwise cross sections were obtained from average parameters and were directly used in neutron simulations. However, it was proven that these infinite-dilute pointwise cross sections cannot capture the energy self-shielding effect, thus leading to an overestimate of reaction rates and usually an underestimate of k{sub eff} owing to the increased absorption rates and increased scattering probability to energy below unresolved resonance region where the capture-to-fission ratio is typically higher. This phenomenon is of great importance especially for problems where there is a considerable flux in unresolved resonance region. Then the probability table method was put forward by Levitt to take energy self-shielding effect into account, which provides probability distribution function of cross sections in unresolved resonance region instead of specific cross section values. The probability distribution function is presented in the form of tables and the mean values are equal to the infinite dilution cross sections. To obtain the temperature dependent cross sections in unresolved energy region, there are two different methods. One is to calculate cross sections directly from average resonance parameters and their statistical distributions , which is effective and precise but time-consuming. The other is to interpolate between pre-generated probability table files, which is much more efficient. In this paper, on-the-fly Doppler broadening of cross sections in unresolved resonance regions based on probability tables interpolation was implemented in RMC codes as a supplement to on-the-fly Doppler broadening for resolved resonance energy region and thermal region, which is based on the integration of Doppler broadening equation. The results of criticality calculation and depletion calculation show that on-the-fly Doppler broadening for the whole energy region is accurate and efficient. (authors)

73 NUCLEAR PHYSICS AND RADIATION PHYSICS↗

Heterogeneous scalable framework for multiphase flows

Two categories of challenges confront the developer of computational spray models: those related to the computation and those related to the physics. Regarding the computation, the trend towards heterogeneous, multi- and many-core platforms will require considerable re-engineering of codes written for the current supercomputing platforms. Regarding the physics, accurate methods for transferring mass, momentum and energy from the dispersed phase onto the carrier fluid grid have so far eluded modelers. Significant challenges also lie at the intersection between these two categories. To be competitive, any physics model must be expressible in a parallel algorithm that performs well on evolving computer platforms. This work created an application based on a software architecture where the physics and software concerns are separated in a way that adds flexibility to both. The develop spray-tracking package includes an application programming interface (API) that abstracts away the platform-dependent parallelization concerns, enabling the scientific programmer to write serial code that the API resolves into parallel processes and threads of execution. The project also developed the infrastructure required to provide similar API s to other application. The API allow object-oriented Fortran applications direct interaction with Trilinos to support memory management of distributed objects in central processing units (CPU) and graphic processing units (GPU) nodes for applications using C++.

71 CLASSICAL AND QUANTUM MECHANICS, GENERAL PHYSIC↗

The OVIS analysis architecture.

This report summarizes the current statistical analysis capability of OVIS and how it works in conjunction with the OVIS data readers and interpolators. It also documents how to extend these capabilities. OVIS is a tool for parallel statistical analysis of sensor data to improve system reliability. Parallelism is achieved using a distributed data model: many sensors on similar components (metaphorically sheep) insert measurements into a series of databases on computers reserved for analyzing the measurements (metaphorically shepherds). Each shepherd node then processes the sheep data stored locally and the results are aggregated across all shepherds. OVIS uses the Visualization Tool Kit (VTK) statistics algorithm class hierarchy to perform analysis of each process's data but avoids VTK's model aggregation stage which uses the Message Passing Interface (MPI); this is because if a single process in an MPI job fails, the entire job will fail. Instead, OVIS uses asynchronous database replication to aggregate statistical models. OVIS has several additional features beyond those present in VTK that, first, accommodate its particular data format and, second, improve the memory and speed of the statistical analyses. First, because many statistical algorithms are multivariate in nature and sensor data is typically univariate, interpolation of data is required to provide simultaneous observations of metrics. Note that in this report, we will refer to a single value obtained from a sensor as a measurement while a collection of multiple sensor values simultaneously present in the system is an observation. A base class for interpolation is provided that abstracts the operation of converting multiple sensor measurements into simultaneous observations. A concrete implementation is provided that performs piecewise constant temporal interpolation of multiple metrics across a single component. Secondly, because calculations may summarize data too large to fit in memory OVIS analyses batches of observations at a time and aggregates these intermediate intra-process models as it goes before storing the final model for inter-process aggregation via database replication. This reduces the memory footprint of the analysis, interpolation, and the database client and server query processing. This also interleaves processing with the disk I/O required to fetch data from the database - also improving speed. This report documents how OVIS performs analyses and how to create additional analysis components that fetch measurements from the database, perform interpolation, or perform operations on streamed observations (such as model updates or assessments). The rest of this section outlines the OVIS analysis algorithm and is followed by sections specific to each subtask. Note that we are limiting our discussion for now to the creation of a model from a set of measurements, and not including the assessment of observations using a model. The same framework can be used for assessment but that use case is not detailed in this report.

97 MATHEMATICS AND COMPUTING↗

Efficient Heterogeneous Execution on Large Multicore and Accelerator Platforms: Case Study Using a Block Tridiagonal Solver

The algorithmic and implementation principles are explored in gainfully exploiting GPU accelerators in conjunction with multicore processors on high-end systems with large numbers of compute nodes, and evaluated in an implementation of a scalable block tridiagonal solver. The accelerator of each compute node is exploited in combination with multicore processors of that node in performing block-level linear algebra operations in the overall, distributed solver algorithm. Optimizations incorporated include: (1) an efficient memory mapping and synchronization interface to minimize data movement, (2) multi-process sharing of the accelerator within a node to obtain balanced load with multicore processors, and (3) an automatic memory management system to efficiently utilize accelerator memory when sub-matrices spill over the limits of device memory. Results are reported from our novel implementation that uses MAGMA and CUBLAS accelerator software systems simultaneously with ACML for multithreaded execution on processors. Overall, using 940 nVidia Tesla X2090 accelerators and 15,040 cores, the best heterogeneous execution delivers a 10.9-fold reduction in run time relative to an already efficient parallel multicore-only baseline implementation that is highly optimized with intra-node and inter-node concurrency and computation-communication overlap. Detailed quantitative results are presented to explain all critical runtime components contributing to hybrid performance.

Park, Alfred J↗

Evaluation of Graph Analytics Frameworks Using the GAP Benchmark Suite

The analysis of connected data is an increasingly important application in high-performance computing. Such analyses can reveal fraudulent patterns in financial transactions, optimize telecommunications networks, predict information flow in social networks, etc. However, the landscape of graph analytics is highly diverse. Graph algorithms stress processor architectures differently, and no one graph can represent all topologies. Consequently, no single approach or framework is expected to be optimal for all graph analytics problems. To help make sense of this diverse landscape, we evaluated four approaches to graph analytics: GraphBLAS, Galois, BGL17, GraphIt; and compare them against hand-tuned implementations that take advantage of hardware features on our test platform. Graph- BLAS formulates graph analytics as sparse linear algebra. Galois provides syntactic constructs for data parallelism over irregular data structures. BGL17 is a generic C++ template library for implementing graph algorithms. GraphIt provides a domain- specific language to describe and optimize graph algorithms. We use the GAP Benchmark Suite to establish baseline performance and guide the side-by-side evaluation of each framework. GAP consists of 30 tests: six graph analytics algorithms (breadth- first search, single-source shortest path, PageRank, betweenness centrality, connected components, and triangle counting) run on five graphs, each with different topological characteristics (e.g., high diameter, skewed degree distribution, high average degree). High-performance reference implementations are included for each benchmark algorithm. Because a graph can be loaded into memory a number of ways (e.g., flat file on disk, compressed sparse format, data frames, retrieved from SQL or NoSQL databases), our evaluation focused on computational performance rather than I/O. Our results show the relative strengths of each framework.

Graph algorithms, Benchmarking, shared-memory prog↗

Large-Scale Modeling of Epileptic Seizures: Scaling Properties of Two Parallel Neuronal Network Simulation Algorithms

Our limited understanding of the relationship between the behavior of individual neurons and large neuronal networks is an important limitation in current epilepsy research and may be one of the main causes of our inadequate ability to treat it. Addressing this problem directly via experiments is impossibly complex; thus, we have been developing and studying medium-large-scale simulations of detailed neuronal networks to guide us. Flexibility in the connection schemas and a complete description of the cortical tissue seem necessary for this purpose. In this paper we examine some of the basic issues encountered in these multiscale simulations. We have determined the detailed behavior of two such simulators on parallel computer systems. The observed memory and computation-time scaling behavior for a distributed memory implementation were very good over the range studied, both in terms of network sizes (2,000 to 400,000 neurons) and processor pool sizes (1 to 256 processors). Our simulations required between a few megabytes and about 150 gigabytes of RAM and lasted between a few minutes and about a week, well within the capability of most multinode clusters. Therefore, simulations of epileptic seizures on networks with millions of cells should be feasible on current supercomputers.

59 BASIC BIOLOGICAL SCIENCES↗

Parallel Multivariate Spatio-Temporal Clustering of Large Ecological Datasets on Hybrid Supercomputers

A proliferation of data from vast networks of remote sensing platforms (satellites, unmanned aircraft systems (UAS), airborne etc.), observational facilities (meteorological, eddy covariance etc.), state-of-the-art sensors, and simulation models offer unprecedented opportunities for scientific discovery. Unsupervised classification is a widely applied data mining approach to derive insights from such data. However, classification of very large data sets is a complex computational problem that requires efficient numerical algorithms and implementations on high performance computing (HPC) platforms. Additionally, increasing power, space, cooling and efficiency requirements has led to the deployment of hybrid supercomputing platforms with complex architectures and memory hierarchies like the Titan system at Oak Ridge National Laboratory. The advent of such accelerated computing architectures offers new challenges and opportunities for big data analytics in general and specifically, large scale cluster analysis in our case. Although there is an existing body of work on parallel cluster analysis, those approaches do not fully meet the needs imposed by the nature and size of our large data sets. Moreover, they had scaling limitations and were mostly limited to traditional distributed memory computing platforms. We present a parallel Multivariate Spatio-Temporal Clustering (MSTC) technique based on k-means cluster analysis that can target hybrid supercomputers like Titan. We developed a hybrid MPI, CUDA and OpenACC implementation that can utilize both CPU and GPU resources on computational nodes. We describe performance results on Titan that demonstrate the scalability and efficacy of our approach in processing large ecological data sets.

Sreepathi, Sarat↗

Parallel Newton-Krylov-Schwarz algorithms for the transonic full potential equation

We study parallel two-level overlapping Schwarz algorithms for solving nonlinear finite element problems, in particular, for the full potential equation of aerodynamics discretized in two dimensions with bilinear elements. The overall algorithm, Newton-Krylov-Schwarz (NKS), employs an inexact finite-difference Newton method and a Krylov space iterative method, with a two-level overlapping Schwarz method as a preconditioner. We demonstrate that NKS, combined with a density upwinding continuation strategy for problems with weak shocks, is robust and, economical for this class of mixed elliptic-hyperbolic nonlinear partial differential equations, with proper specification of several parameters. We study upwinding parameters, inner convergence tolerance, coarse grid density, subdomain overlap, and the level of fill-in in the incomplete factorization, and report their effect on numerical convergence rate, overall execution time, and parallel efficiency on a distributed-memory parallel computer.

Cai, Xiao-Chuan↗

Parallel Conjugate Gradient: Effects of Ordering Strategies, Programming Paradigms, and Architectural Platforms

The Conjugate Gradient (CG) algorithm is perhaps the best-known iterative technique to solve sparse linear systems that are symmetric and positive definite. A sparse matrix-vector multiply (SPMV) usually accounts for most of the floating-point operations within a CG iteration. In this paper, we investigate the effects of various ordering and partitioning strategies on the performance of parallel CG and SPMV using different programming paradigms and architectures. Results show that for this class of applications, ordering significantly improves overall performance, that cache reuse may be more important than reducing communication, and that it is possible to achieve message passing performance using shared memory constructs through careful data ordering and distribution. However, a multi-threaded implementation of CG on the Tera MTA does not require special ordering or partitioning to obtain high efficiency and scalability.

Oliker, Leonid↗

TAO Users Manual (Rev. 3.15)

The Toolkit for Advanced Optimization (TAO) focuses on the development of algorithms and software for the solution of large-scale optimization problems on high-performance architectures. Areas of interest include unconstrained and bound-constrained optimization, nonlinear least squares problems, optimization problems with partial differential equation constraints, and variational inequalities and complementarity constraints. The development of TAO was motivated by the scattered support for parallel computations and the lack of reuse of external toolkits in current optimization software. Our aim is to produce high-quality optimization software for computing environments ranging from workstations and laptops to massively parallel high-performance architectures. Our design decisions are strongly motivated by the challenges inherent in the use of large-scale distributed memory architectures and the reality of working with large, often poorly structured legacy codes for specific applications.

97 MATHEMATICS AND COMPUTING↗

The OMPS Limb Profiler Instrument: Two-Dimensional Retrieval Algorithm

The upcoming Ozone Mapper and Profiler Suite (OMPS), which will be launched on the NPOESS Preparatory Project (NPP) platform in early 2011, will continue monitoring the global distribution of the Earth's middle atmosphere ozone and aerosol. OMPS is composed of three instruments, namely the Total Column Mapper (heritage: TOMS, OMI), the Nadir Profiler (heritage: SBUV) and the Limb Profiler (heritage: SOLSE/LORE, OSIRIS, SCIAMACHY, SAGE III). The ultimate goal of the mission is to better understand and quantify the rate of stratospheric ozone recovery. The focus of the paper will be on the Limb Profiler (LP) instrument. The LP instrument will measure the Earth's limb radiance (which is due to the scattering of solar photons by air molecules, aerosol and Earth surface) in the ultra-violet (UV), visible and near infrared, from 285 to 1000 nm. The LP simultaneously images the whole vertical extent of the Earth's limb through three vertical slits, each covering a vertical tangent height range of 100 km and each horizontally spaced by 250 km in the cross-track direction. Measurements are made every 19 seconds along the orbit track, which corresponds to a distance of about 150km. Several data analysis tools are presently being constructed and tested to retrieve ozone and aerosol vertical distribution from limb radiance measurements. The primary NASA algorithm is based on earlier algorithms developed for the SOLSE/LORE and SAGE III limb scatter missions. All the existing retrieval algorithms rely on a spherical symmetry assumption for the atmosphere structure. While this assumption is reasonable in most of the stratosphere, it is no longer valid in regions of prime scientific interest, such as polar vortex and UTLS regions. The paper will describe a two-dimensional retrieval algorithm whereby the ozone distribution is simultaneously retrieved vertically and horizontally for a whole orbit. The retrieval code relies on (1) a forward 2D Radiative Transfer code (to model limb radiances within a non-uniform atmosphere and evaluate 2D analytical partial derivatives) and (2) an optimal estimator inversion routine. The algorithm uses the typically sparse nature of the kernel matrices as well as fast matrix inversion techniques to allow for fast inversion of limb data with efficient memory management (as was done for MIPAS data processing). While the method has so far only been developed in the context of Single Scatter, the paper will show how the CPU intensive Multiple Scatter modeling can be implemented using parallel CPU processing. Initial results will be presented in terms of retrieved ozone profiles and code performance.

Rault, Didier F.↗

Optimizing the Accelerated Recursive Doubling Algorithm for Block Tridiagonal Systems of Equations

The need to solve block tridiagonal systems with hundreds or thousands of right-hand sides for the same block tridiagonal matrix is common in a variety of disciplines. To meet this need, the Accelerated Recursive Doubling Algorithm was developed. After a right-hand side independent phase, the algorithm allows for the quick, online calculation of solutions for different right-hand sides. In this work, we present methods to optimize the Accelerated Recursive Doubling Algorithm in memory usage and computation time in a hybrid parallelization model. The right-hand side independent phase of the naïve implementation takes ≥ 11/3 the amount of memory required to store the tridiagonal matrix, while our implementation reduces the fraction to ≈ 5/3 . The right-hand side dependent phase of the naïve implementation takes ≥ 6 times the amount of memory required to store the right-hand side, while our implementation reduces the fraction to ≈ 3. The computation time for the independent phase is reduced to ≈ 2/3 times that of the naïve implementation, while the computation time for the dependent phase is reduced to ≈ 5/9 . With increasing numbers of shared-memory threads q on every distributed processing element, we have O(q) theoretical speedup.

97 MATHEMATICS AND COMPUTING↗

Incremental Parallelization of Non-Data-Parallel Programs Using the Charon Message-Passing Library

Message passing is among the most popular techniques for parallelizing scientific programs on distributed-memory architectures. The reasons for its success are wide availability (MPI), efficiency, and full tuning control provided to the programmer. A major drawback, however, is that incremental parallelization, as offered by compiler directives, is not generally possible, because all data structures have to be changed throughout the program simultaneously. Charon remedies this situation through mappings between distributed and non-distributed data. It allows breaking up the parallelization into small steps, guaranteeing correctness at every stage. Several tools are available to help convert legacy codes into high-performance message-passing programs. They usually target data-parallel applications, whose loops carrying most of the work can be distributed among all processors without much dependency analysis. Others do a full dependency analysis and then convert the code virtually automatically. Even more toolkits are available that aid construction from scratch of message passing programs. None, however, allows piecemeal translation of codes with complex data dependencies (i.e. non-data-parallel programs) into message passing codes. The Charon library (available in both C and Fortran) provides incremental parallelization capabilities by linking legacy code arrays with distributed arrays. During the conversion process, non-distributed and distributed arrays exist side by side, and simple mapping functions allow the programmer to switch between the two in any location in the program. Charon also provides wrapper functions that leave the structure of the legacy code intact, but that allow execution on truly distributed data. Finally, the library provides a rich set of communication functions that support virtually all patterns of remote data demands in realistic structured grid scientific programs, including transposition, nearest-neighbor communication, pipelining, gather/scatter, and redistribution. At the end of the conversion process most intermediate Charon function calls will have been removed, the non-distributed arrays will have been deleted, and virtually the only remaining Charon functions calls are the high-level, highly optimized communications. Distribution of the data is under complete control of the programmer, although a wide range of useful distributions is easily available through predefined functions. A crucial aspect of the library is that it does not allocate space for distributed arrays, but accepts programmer-specified memory. This has two major consequences. First, codes parallelized using Charon do not suffer from encapsulation; user data is always directly accessible. This provides high efficiency, and also retains the possibility of using message passing directly for highly irregular communications. Second, non-distributed arrays can be interpreted as (trivial) distributions in the Charon sense, which allows them to be mapped to truly distributed arrays, and vice versa. This is the mechanism that enables incremental parallelization. In this paper we provide a brief introduction of the library and then focus on the actual steps in the parallelization process, using some representative examples from, among others, the NAS Parallel Benchmarks. We show how a complicated two-dimensional pipeline-the prototypical non-data-parallel algorithm- can be constructed with ease. To demonstrate the flexibility of the library, we give examples of the stepwise, efficient parallel implementation of nonlocal boundary conditions common in aircraft simulations, as well as the construction of the sequence of grids required for multigrid.

VanderWijngaart, Rob F.↗